Controversy exists about the optimal amount and source of dietary carbohydrate for managing insulin resistance. Therefore, we compared the effects on insulin sensitivity (SI), pancreatic responsivity (AIRglu) and glucose disposition index (DI=SI×AIRglu) of dietary advice aimed at reducing the amount or altering the source of dietary carbohydrate in subjects with impaired glucose tolerance (IGT). Subjects were randomized to high-carbohydrate–high-glycaemic index (GI) (high-GI, n 11), high-carbohydrate–low-GI (low-GI, n 13), or low-carbohydrate–high-monounsaturated fat (MUFA, n 11) dietary advice, with SI, AIRglu and DI measured using a frequently sampled, intravenous glucose tolerance test before and after 4 months treatment. Carbohydrate and fat intakes and diet GI, respectively, were: high-GI, 53 %, 28 %, 83; low-GI, 55 %, 25 %, 76; MUFA, 47 %, 35 %, 82. Weight changes on each diet differed significantly from each other: high-GI, -0·49 (SEM 0·29) KG; LOW-GI, -0·19 (sem 0·40) kg; MUFA +0·27 (sem 0·45) kg. Blood lipids did not change, but glycated haemoglobin increased significantly on MUFA, 0·02 (sem 0·11) %, relative to low-GI, -0·19 (sem 0·08) %, and high-GI, -0·13 (sem 0·14) %. Diastolic blood pressure fell by 8 mmHg on low-GI relative to MUFA (P=0·038). Although SI and AIRglu did not change significantly, DI, a measure of the ability of β-cells to overcome insulin resistance by increasing insulin secretion, increased on low-GI by >50 % (P=0·02). After adjusting for baseline values, the increase in DI on low-GI, 0·17 (sem 0·07), was significantly greater than those on MUFA, -0·09 (sem 0·08) and high-GI, -0·03 (sem 0·02) (P=0·019). Thus, the long-term effects of altering the source of dietary carbohydrate differ from those of altering the amount. High-carbohydrate–low-GI dietary advice improved β-cell function in subjects with IGT, and may, therefore, be useful in the management of IGT.